Introduction

Selection can be a potent mechanism of phenotypic change in complex eukaryotes, effectively coupling trait values to environmental variables. One superb example, at least in laboratory populations, is the speed and magnitude of phenotypic evolution in Drosophila melanogaster. Indeed, diverse traits are highly responsive to selection by extreme temperature, desiccation, toxic substances, gravitrophism, starvation, spontaneous mortality, and reproductive effort, among others (Hoffmann and Parsons, 1993; Ricker and Hirsch, 1998; Djawdan et al, 1998; Promislov et al, 1998; Bryant and Reed, 1999; Singh and Singh, 2001; Hoffmann et al, 2003a). For at least some of these agents, moreover, relaxation of selection causes the resultant phenotypes to revert to their pre-existing levels, suggesting that ongoing selection may be necessary to maintain the values of such traits (Teotonio and Rose, 2001; Teotonio et al, 2002). Thus, these findings are consistent with all of the classical conditions for selection (ie, genetic encoding of relevant phenotypes, variation in the encoding genes, differential fitness under selection) persisting at sufficiently high levels in laboratory populations of D. melanogaster.

In nature, by contrast, the complex, dynamic, and unstable interplay of diverse environmental, demographic, and genetic variables can undermine the directionality, if not the potency of natural selection. That despite this dynamism selection can create monumental phenotypic diversification in nature is self-evident. But are such instances of diversification extremely unlikely, with reversals in selection pressure or demographic interference (eg, swamping via migration) typically eroding nascent differences before they have an opportunity to accumulate and fix? D. melanogaster is both small and highly mobile, which prospectively exposes it to diverse, if not contrasting, selection pressures on a micro scale and enhances gene flow among local populations. Thus, sustained unidirectional selection in D. melanogaster may be unlikely. Most conspicuous exceptions to this expectation involve large-scale clines (Alonso-Moraga et al, 1988; David et al, 1989; James et al, 1997; Robinson et al, 2000; Huey et al, 2001; Verrelli and Eanes, 2001; Duvernell et al, 2003; Hoffmann et al, 2003b; Gilchrist et al, 2004), which may override local heterogeneity in selection pressures, or peculiar instances of partial isolation (eg, wine cellar populations).

By contrast, in Lower Nahal Oren, Mt Carmel, Israel (‘Evolution canyon’), populations living in distinctive microclimates have diverged in multiple traits despite minimal spatial separation. The opposite slopes of this canyon show strong abiotic contrasts that are consequential for species composition and population genetic structure in diverse organisms, including several Drosophila species (Nevo, 1997, 2001; Nevo et al, 1998; Harry et al, 1999; Pavlicek et al, 2003). The D. melanogaster populations on the slopes, separated by 100 and 400 m at the bottom and top, respectively, experience markedly different environments due to the higher illumination on the south-facing slope (SFS) than on the north-facing slope (NFS) (Pavlicek et al, 2003). The slopes also differ in temperature and aridity: NFS has comparatively lush vegetation of European origin, whereas the SFS is an open Park Forest or Xeric Savanna, primarily of African and Asian origin. The Drosophila populations in the canyon differ in habitat choice, thermotolerance and desiccation resistance, and life-history traits (Nevo et al, 1998; Rashkovetsky et al, 2000; Iliadi et al, 2001; Lupu et al, 2004), all corresponding to the prevailing microclimate. These populations also differ in sexual behavior, including mate choice (Korol et al, 2000; Iliadi et al, 2001; Drake et al, 2005). This remarkable divergence has evolved despite an interslope distance much smaller than the daily dispersal capability of Drosophila (Coyne and Milstead, 1987).

Here we ask: Were these first reports of interpopulation differences in ‘Evolution Canyon’ Drosophila an unrepresentative snapshot of a temporary deviation from homogeneity, or are these differences stable despite year-to-year and seasonal variation in environmental conditions and potential interslope migration? Our results support the latter.

Materials and methods

During July–October 1997, 1999, and 2000, we collected wild female D. melanogaster from yeasted banana bait at the two mid-stations (90 m above sea level) on the NFS and SFS of Lower Nahal Oren canyon (Mount Carmel, Israel). Isofemale lines were established from each female inseminated in nature and maintained under standard conditions (25°C; approximately 40% mean relative humidity; standard cornmeal–sugar–agar medium). Synthetic populations were established for each slope and year by combining 10 flies of each sex from 25 isofemale lines in a population cage, and maintained under random mating for 72 (1997 collection), 24 (1999), and 12 (2000) nonoverlapping generations. To examine the impact of inbreeding, lines established from the 1997 synthetic population were sib-mated for eight generations.

Thermotolerance measurements

Adult flies were transferred to fresh bottles and allowed to oviposit, and then cleared from the bottles. We collected the first brood of flies eclosing during an 18-h window beginning at 1800 h. After CO2 anesthesia, these flies were sorted by sex into groups of 20, and each group transferred to 22 × 95 mm glass vials containing 8 ml of medium. After two additional days at 25°C, these vials were stoppered with cotton plugs, inverted, and fastened to plastic racks, which were submerged in circulating water baths (GFL-1083, Gesellschaft fur Labortechnic mbH, Burgwedel) regulated within ±0.3°C of the temperatures indicated below. Survival was scored 24 h after heat treatment as the proportion of flies in a vial exhibiting any response to touching with a fine paint brush. This scheme was based on previous Drosophila thermotolerance studies (Loeschcke and Krebs, 1997; Krebs and Feder, 1998; Bettencourt et al, 1999). Heat treatments were:

  1. a)

    Heat shock only (HS): 38.5 (for 1997 and 1999 populations) or 39 (for 2000 populations) ±0.3°C for 50, 60, or 70 min;

  2. b)

    Heat pretreatment (PT) at 36°C for 1 h and 25°C for 1 h preceding heat shock as described above (PT+HS).

Comparisons and preliminary experimentation

Often, eggs deposited on the same day will yield adults that eclose over several days; we term adults eclosing on each day a brood. Preliminary studies revealed that brood affected thermotolerance (see also Sorensen and Loeschcke, 2004). Hence, except where noted, experiments used only the first brood (excepting rare rapidly eclosing adults) eclosing from any day's egg deposition.

We compared thermotolerance of synthetic populations, isofemale lines, and inbred lines from each slope in a factorial design with replicated tests. The factors included were: ‘year’ (1997, 1999, and 2000), ‘population’ (SFS and NFS), and ‘treatment’ (PT+HS and HS only).

Data analysis

Statistical tools including ANOVA, and log-linear analysis were employed for data analysis using Statistica software package (StatSoft, 1996).

Results

Thermotolerance: comparing synthetic populations

Slope, sex, year of collection, duration of heat shock, and pretreatment all affected the survival of heat shock (see Figure 1). Not unexpectedly, survival was inversely related to the duration of heat shock. Consistent with numerous prior reports, pretreatment increased the mean thermotolerance in every comparison (2 years × 3 temperatures × 2 sexes × 2 slopes), significantly so in 17 of 24 cases (Table 1).

Figure 1
figure 1

Basal (gray columns) and inducible (white columns) thermotolerance of Drosophila from the opposite slopes of Nahal Oren canyon. Heat shock was at 38.5°C and pretreatment was at 36°C for 60 min.

Table 1 Effect of pretreatment on the percentage of Drosophila from NFS and SFS surviving heat shock of 38.5°C

Our major interest is in the interslope differences in thermotolerance. In each of the 24 SFS vs NFS comparisons (2 years × 3 temperatures × 2 sexes × ±pretreatment), the mean thermotolerance for the SFS sample exceeds the mean thermotolerance for the NFS sample (see Table 2).

Table 2 Effect of slope of origin on the percentage of Drosophila from the NFS and SFS surviving heat shock of 38.5°C

These differences in means, moreover, were statistically significant in 11 of the 12 comparisons for acquired thermotolerance. The differences were less frequently significant for basal thermotolerance, where the mean thermotolerance was small relative to the sampling error. The mean change in thermotolerance between pretreated and unpretreated flies was greater for SFS than for NFS flies in 11 of 12 comparisons; overall, this difference was significant (Table 3). Repetition of the experiment with single heat shock duration, pretreated flies only, and flies collected in 2000 yielded similar outcomes (Tables 3, 4).

Table 3 Log-linear analysis of thermotolerance in D. melanogaster populations derived from opposite slopes of ‘Evolution Canyon’
Table 4 Analysis of the effects of lines, flies' origin and sex on inducible thermotolerance for Drosophila collected in 2000

Thermotolerance: comparing isofemale lines

Repetition of the above studies with separate isofemale lines founded from flies collected in 2000 (Figure 2) revealed patterns of variation similar to those evident in synthetic populations. Log-linear analysis revealed that the founders' slope of origin was the most significant factor affecting survivorship after pretreatment. As before, SFS flies were more tolerant than NFS flies (Figure 2). Sex had a lesser impact than for synthetic populations, which was significant only for SFS lines (Table 4). In addition, separate lines founded from parents from the same slope varied significantly in thermotolerance, and more so in the lines from the SFS. The greater variation in the SFS lines is evident from the ratio of two χ2 statistics, which is distributed asymptotically as Fisher's F-statistics with corresponding degrees of freedom (F7,5=154.71/24.33=6.36, P=0.029). Between-slope differences were principally due to the large differences among female flies.

Figure 2
figure 2

Thermotolerance of isofemale lines from the opposite slopes. Inset: Grand means by slope and sex for the isofemale lines. Means are plotted ±1 SE. Heat shock was at 39°C.

Strongly inbred lines

Strongly inbred lines exhibited essentially the same patterns of variation in thermotolerance that synthetic populations and isofemale lines displayed, except that the average thermotolerances were dramatically lower after comparable treatments in the strongly inbred lines (Table 5, Figure 3). Importantly, thermotolerances of inbred flies were more variable for lines founded from the SFS than from the NFS.

Table 5 Analysis of the effects of flies' origin, lines, sex and type of treatment on inbred lines' thermotolerance
Figure 3
figure 3

Comparison of thermotolerance in inbred and outbred lines of Drosophila from the opposite slopes of Nahal Oren canyon.

Discussion

The initial reports of inter-slope differentiation in ‘Evolution Canyon’ have elicited considerable controversy. Drosophila adults are able to disperse long distances (10–15 km) overnight (Coyne and Milstead, 1987; Coyne et al, 1987). Therefore, that slope-specific adaptive gene complexes could evolve at all and escape recombinational collapse is perplexing, even to the authors, and has prompted numerous subsequent investigations, which themselves are controversial. For example, if the populations on the two slopes are indeed distinct on a sustained basis, their genes should diverge in sequence. Indeed, our recent estimates based on microsatellite markers (Michalak et al, 2001) revealed a substantial interslope differentiation for microsatellites in D. melanogaster as great as between it and its sibling species, D. simulans, and indicated that gene flow should be rather restricted among the slopes. Schlotterer and Agis (2002) and Colson (2002), by contrast, examining many of the same microsatellites in flies collected at nearly identical times, found scant genetic differentiation. Additionally, NFS and SFS populations sampled in 1995 did not differ at specific loci for the Acp gene family (Panhuis et al, 2003). Interslope genetic differentiation in D. melanogaster derived from ‘Evolution Canyon’ was revealed in our recent study of the period gene known to affect sexual behavior. Variants of the (Thr–Gly)n repeat of the period gene, n=17 and n=20, which are abundant in natural populations of D. melanogaster in Africa and Europe (Kyriacou et al, 1996; Sawyer et al, 1997), were found to predominate in the Canyon. A noteworthy fact is that the less abundant ‘European’ allele (n=20) occurred on the NFS about three-fold compared to the SFS (Zamorzaeva et al, 2005). These reports could be reconciled if the ‘Evolution Canyon’ Drosophila populations were undergoing dynamic demographic and environmental change, which is certainly possible for small insects living in such a variable environment. Thus, the Introduction asked: Were the first reports of interpopulation differences in ‘Evolution Canyon’ Drosophila an unrepresentative snapshot of a temporary deviation from homogeneity, or are these differences stable despite year-to-year and seasonal variation in environmental conditions and potential interslope migration?

Our principal finding is that the difference in thermotolerance between flies from the two slopes of the canyon is ongoing and robust. Greater thermotolerance in SFS Drosophila than in NFS slope Drosophila is now evident for flies collected in 1997, 1999, and 2000. These differences, moreover, are in both basal thermotolerance and inducible thermotolerance, and appear in synthetic populations, isofemale lines, and inbred lines. Drosophila from the two slopes also differ in pre-adult viability and developmental time (Rashkovetsky et al, 2000). Since the original study, interslope differences in habitat choice (Nevo et al, 1998) and mating preference (Korol et al, 2000; Iliadi et al, 2001; Drake et al, 2005) have also come to light. At least in phenotypes reported here, the NFS and SFS populations differ.

These differences may be related to adaptation to the contrasting environmental regimes prevailing on the two slopes (Pavlicek et al, 2003). The difference in inducible thermotolerance, moreover, has a candidate genetic basis. The inducible molecular chaperone Hsp70 is responsible for a substantial portion of inducible thermotolerance. In the ‘Evolution Canyon’ populations, naturally occurring P transposable elements disrupt the proximal promoters of at least two of the five Hsp70-encoding genes. Such disruptions can reduce hsp70 mRNA transcription, Hsp70 protein levels, and thermotolerance (Lerman et al, 2003; Lerman and Feder, 2004). In the ‘Evolution Canyon’ Drosophila, hsp70 alleles in which the P element is present or absent segregate in a balanced polymorphism, but at higher allelic frequencies in the NFS population than in the SFS population. This pattern is consistent with the lesser thermotolerance of the NFS population.

Above we show that strong inbreeding reduces thermotolerance. Thus, an alternative explanation is that the inter-slope differences in thermotolerance reflect differing magnitudes of inbreeding on the two slopes. A definitive portrait of the demographic and environmental dynamism that Drosophila undergo in ‘Evolution Canyon’ may need to await the development of truly miniaturized equipment that can report flies' environment, position, and with which other flies they interact.

An additional explanation of the interslope differentiation is that, although Drosophila can travel long distances (Coyne and Milstead, 1987; Coyne et al, 1987), those in ‘Evolution Canyon’ do not. That is, either migration in Nahal Oren canyon is much lower than usually thought for such small distances or there is significant deviation from common simple population-genetic assumptions (ie, panmixia, random dispersal, and weak-to-moderate selection), or all these deviations work together (Korol et al, 2000; Iliadi et al, 2001). Laboratory comparisons of migratory activity between flies from the canyon and a population collected from an open forest park on the Golan Heights (Iliadi et al, 2002) revealed no differences, suggesting that the ‘Evolution Canyon’ flies are not atypical migrators.

A final possible explanation of the discrepancy in the results could be that differential selection initiates the creation of alternative gene complexes (haplotypes) for loci affecting the selected traits on the two slopes, and that their maintenance in the face of gene flow/recombination is due to selection facilitated by certain habitat choice (Nevo et al, 1998) and assortative mating (Korol et al, 2000; Iliadi et al, 2001; Singh et al, 2005). In this scenario, adaptive differentiation can withstand the disruptive effects of migration and recombination. Such adaptive differentiation, however, would not necessarily be accompanied by differentiation of selectively neutral markers, unless the latter are in linkage disequilibrium with selected loci. This last condition can also persist despite migration, but only under tight linkage and strong selection. In a number of Drosophila genes, linkage disequilibrium decays within a few kilobases (kb), or even within 1 kb (Langley et al, 2000). Therefore, differentiation of adaptive trait complexes seems to provide better evidence for interslope differential selection than that displayed by genetic distances estimated using molecular markers.

In conclusion, initial responses to reports of evolved differences between Drosophila populations on the two sides of ‘Evolution Canyon’ were appropriately skeptical. The present study clearly demonstrates that these differences were not a one-time occurrence, but either persist or re- evolve over multiple years. In principle, moreover, the microevolution of thermotolerance should not be confined to ‘Evolution Canyon’, but demonstrable wherever similar microclimatic gradients exist. Testing this prediction might well elucidate the specific evolutionary mechanisms that have given rise to the differentiation in ‘Evolution Canyon’.